& The Notion of ‘Phonology’ inDyslexia Research:Cognitivism}and BeyondPer Henning Uppstad and Finn Egil T�nnessen

Phonology has been a central concept in the scientific study ofdyslexia over the past decades. Despite its central position,however, it is a concept with no precise definition or status. Thepresent article investigates the notion of ‘phonology’ in the traditionof cognitive psychology. An attempt is made to characterize thebasic assumptions of the phonological approach to dyslexia and toevaluate these assumptions on the basis of commonly acceptedstandards of empirical science. First, the core assumptions ofphonological awareness are outlined and discussed. Second, theposition of Paula Tallal is presented and discussed in order to shedlight on an attempt to stretch the cognitive-psychological notion of‘phonology’ towards auditory and perceptual aspects. Both the coreassumptions and Tallal’s position are rejected as unfortunate, albeitfor different reasons. Third, the outcome of this discussion is asearch for what is referred to as a ‘vulnerable theory’ within thisfield. The present article claims that phonological descriptions mustbe based on observable linguistic behaviour, so that hypotheses canbe falsified by data. Consequently, definitions of ‘dyslexia’ must bebased on symptoms; causal aspects should not be included. In fact,we claim that causal aspects, such as ‘phonological deficit’, bothexclude other causal hypotheses and lead to circular reasoning. Ifwe are to use terms such as ‘phonology’ and ‘phoneme’ in dyslexiaresearch, we must have more precise operationalizations of them.

Keywords: phonology; connectionism; phonological deficit; circularity; definition; phono-logical awareness

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by NORA - Norwegian Open Research Archives

INTRODUCTION

Until about the end of the 1960s, a common view of dyslexia was that itarose as a consequence of sensory or perceptual anomalies in the visualsystem. In the early 1970s, the focus moved to cognitive and linguistic

factors}in particular, phonological factors. At about the same time, we also sawa rekindling of interest in biological factors, which had been a central concern atthe very start of research on dyslexia (Miles & Miles, 1990). In recent years, PaulaTallal has brought the focus back to sensory and perceptual factors, but this timein the auditive system. Her work thus represents a break with tradition}but nota complete break, because she claims that phonological factors are still relevant(cf. Brady, Scarborough, & Shankweiler, 1996).

The present article consists of three major parts. First, we look into thephonological-deficit hypothesis in order to present and discuss the characteristicsof the construction of phonology in this tradition. The second part examines thework of Paula Tallal, which constitutes a contemporary alternative to thepredominant phonological-deficit hypothesis. In this second part, we wish topresent and discuss the findings made, the theories used to explain the data andthe various claims made by Tallal and her co-workers. The motivation for doingso is the search for conceptual clarity, because we claim that the notion of‘phonology’ is highly unclear both in the original hypothesis and in Tallal’sversion of it. The third part investigates the notion of ‘phonology’ beyondcognitivism. In this part, we do not give prominence to any one direction ofphonology}several can in fact be considered useful. Instead, we focus on thebasic prerequisites for such approaches to be empirically sound.

THE PHONOLOGICAL-DEFICIT HYPOTHESIS

The notion of ‘dyslexia’ has been defined in different ways over the past decades.However, notions like ‘phonology’, ‘awareness’ and ‘phoneme’ have been centralin the dominating approaches to dyslexia. In short, it can be summarized that‘phonology’ denotes the sound system, ‘phoneme’ denotes the basic unit in thissystem and ‘awareness’ is the means by which the individual can gain access tothe sound system.

It should not be too controversial to claim that most dyslexia research in the past30 years has built on a particular strand of linguistic theory. The linguists John C.Gumperz and Stephen C. Levinson describe how, in the 1960s, linguistics andpsychology contributed to the rise of cognitive sciences ‘which favoured a strongemphasis on the commonality of human cognition and its basis on human geneticendowment’ (1996, p. 3). In cognitive sciences, the dominant linguistic strand hasbeen generative grammar, a theory which gives priority to underlying linguisticstructures through a distinction between competence (innate language faculty) andperformance (speech). According to this distinction, competence carries theuniversal commonalities while performance is the more or less random realizationof these structures in the particular language and situation (Chomsky, 1975). Thisdistinction is also at the core of what is called autonomous linguistics:

Just as autonomous linguistics distinguishes between a speaker’s purely linguisticknowledge, determined by the language faculty, and his non-linguistic knowledge,

derived from pragmatic competence and the conceptual system, so autonomousphonology splits off the act of speech as an articulatory, acoustic, and perceptual eventfrom the abstract linguistic system which is claimed to underlie the physical data.(Taylor, 1991, p. 28)

The belief in the existence of an underlying system along these lines is firm indyslexia research: ‘We earlier defined phonological coding as the ability to usespeech codes to represent information in the form of words and parts of words.’(Vellutino, Fletcher, Snowling, & Scanlon, 2004). However, as we are not alone inthinking, this approach is unfortunate if the focus is to be kept on empiricalscience:

Faith is a beautiful thing, which a non-believer can only regard with awe. At the timeof writing, that seems to be how matters stand. There are many who share Chomsky’sfaith, and for them ‘The research program of modern linguistics’ (Chomsky 1988: Ch. 2)must without arrogance be his program. Others cannot bring themselves to accept thatit has empirical content. ‘C’est magnifique, mais peut-etre ce n’est pas la linguistique.’But all they can do is watch and wait. (Matthews, 1993, p. 252)

First, autonomous linguistics presupposes the existence of structures which aretoo abstract to be falsified by data}in other words, structures which cannot betreated as hypotheses (see discussion below). Second, the focus on underlyingstructures creates a huge amount of metaperspectives with a highly uncertainstatus. As an example, the linguistic descriptions of semantics, phonology,morphology and syntax are awarded the status of ‘processes’ in cognitive flow-charts (cf. Vellutino et al., 2004, p. 4) without consideration of their function inlinguistics. The fallacy of equating linguistic descriptions with mental processeswas committed early on in Chomsky’s authorship; to some extent alreadybetween Syntactic Structures (1957) and Aspects (1965). We consider this positionto represent naıve realism.

Our conception of linguistic behavior is biased by a tendency to treat processes,activities and conditions on them in terms of object-like, static, autonomous andpermanent structures, i.e., as if they shared such properties with written characters,words, texts, pictures and images. Though my discussion will be confined to linguistics,I am well aware that the same type of bias can be found in many other sciences.(Linell, 1982, p. 1)

As part of the turn towards linguistics and phonology in dyslexia research,scholars subscribing to the phonological-deficit hypothesis have adopted andreinforced central assumptions of orthodox generative grammar (see Table 1).This involves assuming the position of autonomous linguistics. However, thisapproach is problematic from an empirical point of view. Because of thedistinction made between a stable and innate system (competence) and avariable, random use of this system (performance), every observable phenom-enon is subordinated the abstract system by phonological rules. As aconsequence, such rules obstruct the endeavour to meet high standards ofempirical science. This is because priority is given to an abstract system with anaxiomatic base, and the phonological rules formulate how the abstract system canaccount for variation in speech. In other words: it is hypothesized that there is a

deficit in the phonological system, but it is difficult to see how this hypothesiscan be falsified. In this sense, a specific conception of phonology becomesan a priori entity which is left beyond discussion. This is remarkable consideringthe development of phonological theory in linguistic circles over the sameperiod of time}a period characterized by major discoveries and advances(Horne, 2000).

The notions of ‘phonology’ and ‘awareness’

It can be argued that the turn towards phonology in dyslexia research wasinfluenced by assumptions from the linguistic tradition in the joint venture oflinguistics and cognitive psychology in the 1960s. In retrospect, it is evident thatthe turn towards ‘phonology’ was not accompanied by a thorough reflection ofthe adopted assumptions. There was simply no group of critical researchersfocusing on the linguistic assumptions. Alvin Liberman is for many scholars anearly linguistic alibi for reading and dyslexia research, however, his view of therelation between spoken and written language (Liberman et al., 1989) reflectslittle more than traditional and questionable assumptions in the tradition oflinguistics (Linell, 1982; Wengelin, 2002, Uppstad, 2005).

The lack of reflection on the notion of ‘phonology’ may turn out to have direconsequences, not least with regard to the extensive theory-building that has

Table 1. Overview of basic assumptions and common inferences underpinning the theoryof ‘phonological awareness’

Phonology is linguistic, written language is notInferences

The lexicon can be accessed only by speechPhonological structure is directly relevant to written language, but not vice versa

Phonology is natural, written language is an artifactInferences

Speech is a matter of ontogenesis, written language is simply an artifactSpeech is part of the human being, written language is external

Phonology means segmental, linear structureInferences

Prosody is peripheral, not part of phonological structureSound units are linearly ordered, discrete or notSegmental structure is universal

Phonology is platonicInferences

Phonology is abstract, distant from the phonetic, acoustic pronunciationPhonology must be accessed when learning to readPhonology can be accessed only by awareness of phonological units

Phonology is highly representationalInferences

There is a specific mental level where a phonological structure with psychological realityand communicative relevance can be identified and describedThe representation is distant, and qualitatively different, from speech

been undertaken on the basis of the phonological-deficit hypothesis. What ismore, there is a general vagueness in the research field concerning what‘phonology’ means. Even the famous linguist Roman Jakobson refused to use theterm ‘phonology’ back in the late 1970s, because of the conceptual mess(Jakobson, Taylor, & Waugh, 2002). And yet the confusion referred to by Jakobsonis minor compared with the vagueness of the notion of ‘phonology’ identifiableover the past decades in the field of reading research. The Jakobsonian objectionaddresses the rejection by generative grammar of the structural view ofdistinctiveness, and therefore it amounts to precise criticism of a competingtheory. In reading research, the maintenance of the notion is inconsistent,denoting partly letter sounds, the sound system, sounds in general, a languagemodule in the brain, as well as mental units of sound. The outcome of thisinconsistency is that the notion of ‘phonology’ has been fully encompassed by thenotion of phonological awareness. This claim can be made on the basis of thefollowing assumptions:

(1) Phonological structure is what is accessible by awareness: Because phonologyis abstract, it is different and distant from the acoustic signal and theacoustic shape of speech. Phonological structure can therefore be accessedonly by awareness of units. According to this assumption, speech mustbe phonetic in general but the units which one can be aware of (phonemes)are phonological.

(2) The distinction between the notions of ‘phonological’ and ‘phonetic’ is notoperationalized: The problem is that ‘phonological awareness’ is an extremelyvague concept and is therefore of little use in the maintenance of thephonetics–phonology distinction.

(3) Awareness is a necessary and sufficient condition for category membership: Thismeans that there are no categories other than those we can be aware of.The phonological-deficit hypothesis focuses on awareness of segments, anduses this awareness to represent phonological skills. There is no tradition inthe research field of dyslexia for extensive, phonological analysis of wordsand utterances. Phonological rules are relevant only to a certain extent,and they are used mainly to describe aspects of awareness. According tothis assumption, awareness is a necessary and sufficient condition forthe evaluation of phonological skills. A second problem related to thisassumption is the belief that awareness can be measured independently fromother skills.

The arguments concerning ‘phonological’ and ‘phonetic’ may seem like hair-splitting to some scholars. In our view, however, they are of great importance,mainly because the basic assumptions about phonology inherent in thephonological-deficit hypothesis require that a distinction be made between thetwo terms. This is because the basic theory underpinning the phonological-deficithypothesis is based on autonomous linguistics, and the hypothesis presupposestwo distinct levels of explanation: a phonological one and a phonetic one(speech). In the research tradition focusing on ‘phonological awareness’, it ishardly ever shown how to move between these two levels of description byphonological rules which are teleological and normative. What is more, few canperform such analyses in any consistent way, and even fewer have seen anyreason to do so. This is alarming, because the use of phonological rules is the only

way to show the connection between the two levels in autonomous linguistics. Toput it short: autonomous linguistics in general can be questioned from anempirical perspective, but autonomous phonology without phonological rules isnothing less than a disaster. However, in some of the directions taken bygenerative phonology in the last decade, there is also a blurring of the distinctionas to what is ‘phonological’ and ‘phonetic’, respectively (Kager, 1999), but herethe blurring is intentional and founded on basic assumptions about acquisition.In the phonological-deficit hypothesis, though, the blurring is accidental andprojects a flaw in the conceptual system.

If we take a look at Vellutino’s work from 1979, we find that his intentions areambitious:

The primary purpose is to examine the theoretical foundations of these conceptualiza-tions, closely scrutinizing their logical consistency, their empirical validity, their overallproductivity. A special effort is made to discuss the methodology and conclusionsdrawn from the studies reviewed, as well as to analyze their conceptual bases.(Vellutino, 1979, p. 3)

Considering the fact that this book introduces the ‘linguistic turn’ in dyslexiaresearch, it is highly remarkable that none of the intentions quoted above aims atthe core concepts of the new paradigm}the notions of ‘linguistic’ and‘phonology’. That is why we will make an effort in this part of the article tooutline how the notions of ‘phonology’ and ‘linguistic’ are established in thewritings of some central proponents of the phonological-deficit hypothesis. Thepurpose of doing so is primarily to include the notion of ‘phonology’ in thehypothetical construct of the phonological-deficit hypothesis.

The foundation of the phonological-deficit hypothesis

The notion of ‘phonology’ in dyslexia research is linked to a specific conceptionof ‘linguistic’. Since the phonological-deficit hypothesis is an explanation of howthe acquisition of written language is hampered by a deficit in the ability forspoken language, the notion of ‘linguistic’ is restricted to spoken language.Written language is only ‘linguistic’ in the sense that it is dependent on spokenlanguage skills. (Liberman et al., 1989; Vellutino, 1979; Vellutino et al., 2004).Throughout Vellutino’s book on dyslexia (1979), the notion of ‘linguistic’ isreserved for speech. Explicitly, the same position is held in the 2004 article ofVellutino et al.:

Visual coding processes, broadly defined, refer to sensory and higher-level visualiza-tion processes that facilitate storage of representations defining the visual attributes ofenvironmental stimuli, including the graphic symbols used to represent written words.Linguistic coding processes refer to processes that facilitate language acquisitionand the use of language for coding, storing and retrieving information. (Vellutino et al.,2004, p. 4)

The distinction between ‘visual’ and ‘linguistic’ may well be because thelinguistic turn was an antithesis to earlier hypotheses implying visual deficits.As a consequence, the notion of ‘linguistic’ denotes speech, and written languageis simply considered as a representation of speech. Paradoxically, in this waywritten language}being parasitic}can be ‘phonological’, but not ‘linguistic’ by

the fact that ‘phonology’ seems to denote a structure of spoken language relevantto written language while ‘linguistic’ denotes language ability in spokenlanguage: ‘Written words are encoded (symbolized) representations of spokenwords, and spoken words are encoded representations of environmentalexperiences and entities.’ (Vellutino et al., 2004, p. 3).

In our view, it is clear that the core assumptions of orthodox generativephonology (Chomsky & Halle, 1968) are adopted by proponents of thephonological-deficit hypothesis, (Churchland, 1986; Gardner, 1985) and thatthese assumptions are maintained in an unfortunate way (see Table 1). In hisarticle (Ohman, 1979) in a Festschrift dedicated to the engineer and phoneticianGunnar Fant, the phonetician Sven Ohman takes Fant’s contributions tolinguistics as his starting-point in discussing mainly how speech technologyenables the researcher to advance beyond subjective introspection. According toOhman’s view, the study of speech has historically been restricted by the lack oftechnology for studying the acoustic signal. Therefore, the traditional analysis ofspeech ‘was derived from a subjective, introspective analysis of what one does inproducing the sounds of an utterance’ (p. xvii). According to him, it is thereforenecessary to confront what we ‘know’ (from subjective introspection) aboutspeech with what the study of the acoustic signal may tell us:

The point is rather that there is a stage in the theoretical analysis of the linguisticstructure of the acoustic speech signal, beyond which unaided auditory perception isincapable of taking us. And it is beyond this stage that we need to reach in order to beable to assess the soundness of some of our most deeply seated theoretical principles.(Ohman, 1979, p. xviii)

Moreover, such a study is necessary in order to reflect on where the conceptionof sound structure comes from. Among what Ohman calls ‘our most deeplyseated theoretical principles’ he highlights three traditional conceptions of soundstructure to be challenged by modern speech technology:

(1) The discreteness principle says that an utterance of spoken language is alwaysmade up of linguistically-indivisible building blocks drawn from a fixed andfinite inventory of types of such blocks [>the compositional view oflanguage].

(2) The linearity principle dictates that these building blocks be always arrangedin a linear series along the time axis.

(3) The principle of articulatory posturality claims that the building blocks of anutterance are best understood as certain postures of the organs of speech.(Ohman, 1979, p. xviii, italics mine)

The major discoveries made in phonology over the past three decades haveled to the above principles all being rejected as unfortunate (cf. Horne, 2000).The advances made in the field of prosody has shown that sound structurecan be better described as acoustic features on different tiers that are associatedwith a segmental tier in a dynamic way (rejection of principle 1 above). Thesame discoveries in prosody has shown that the arrangement of features ismore nuanced than the above mentioned principle 2, and modern phonologyis therefore often referred to as non-linear phonology. The same discoverieshighlight the importance of the characteristics of the dynamic acoustic signal,

while aspects of articulation are not focused in phonological description(principle 3).

This focus on prosodic research is partly due to the fact that developments in speechtechnology have made it possible to examine the acoustic parameters associated withprosodic phenomena (in particular fundamental frequency and duration) to an extentwhich has not been possible in other domains of speech research. It is also due to thefact that significant theoretical advances in linguistics and phonetics have been madeduring this time which have made it possible to obtain a better understanding of howprosodic parameters function in expressing different kinds of meaning in the languagesof the world. (Horne, 2000, p. 1)

Notwithstanding this, the scientific study of dyslexia has maintained theprinciples referred to by Ohman in a quite dogmatic way. This situation meansthat the theoretical principles underpinning the phonological-deficit hypothesisseem highly questionable. In order to make empirical progress, the proponents ofphonological awareness must with precision elaborate what is denoted by theconcepts of both ‘phonology’ and ‘linguistic’.

Implications for application

From the theoretical principles mentioned in the previous paragraph asunderpinning the phonological-deficit hypothesis, some features of appli-cation can be outlined. Every generalization of this kind will be controversial,because the theorists involved certainly do add important reservationsand nuances. Nevertheless, the generalization presented below is probablyrepresentative of the ‘period of normal science’ for this paradigm (Kuhn,1970).

The main problem in dyslexia research is that the phonological theory has notbeen maintained in a way enabling the basic assumptions to be continuouslyexposed to falsification. A second set of problems, which can be derived from thefirst, originate from an unreflected conceptual system. In the situation obtaining,the label ‘phonology’ may be applied to anything related to sound in connectionwith speech and writing; whether a given sound is phonetic or phonological maybe left to the author’s ability and desire to remain consistent throughout a paper,and the outcome of this is that the notion of ‘phoneme’ is the closest one gets tomake explicit what is meant by ‘phonology’.

Problem: the main argument in phonological awareness can be described ascircular

The main argument for the theory of phonological awareness is the highcorrelation between awareness of sound segments and reading performance(Vellutino et al., 2004). This finding has been replicated so often that it isconsidered to reflect a causal relationship. However, two conditions have to bemet for this to be the case, namely: (1) that the structure of speech is strictlysegmental (from an ontological perspective); and (2) that writing is parasitic onspeech, i.e. that written language is considered as simply a representation ofspoken language. The latter condition rests on the idea that ‘Written words areencoded (symbolized) representations of spoken words’ (Vellutino et al., 2004,p. 3). Both conditions can be considered to be assumed in the so-called

alphabetical principle. In fact, (1) has been strongly questioned in phonologicaltheory (Horne, 2000) and (2)}the relation of spoken and written language}hasnever been treated as a true hypothesis. In every other respect, the relationshipbetween awareness and reading performance has strong implications not ofcausality but of circularity: it would seem to amount to saying ‘those whosucceeded, succeeded’ and ‘those who did not succeed, did not succeed’. In otherwords, there is shown to be a high correlation between the level of the sameproperty in the same individual at two points in time, regardless of whether thestudies involved are experimental or interventional. What tends to happen is that‘phonological tests’ are conducted on pre-school children. Later on, when thesame children have entered school, they are given reading tasks which focusprimarily on the ‘phonological part’ of reading. No wonder that the correlation ishigh! For the purpose of illustration, a corresponding situation could beimagined if we, say, assessed IQ in pre-school children and then compared theresults obtained with results obtained from the same children after they hadentered school. The obvious origin of the circularity is the segmental analysis ofspeech, which fits in with the segmental structure of writing. With regard to thefirst point above, relating to the segmental structure of speech, modernphonology has shown how phonological structure does not favour the segmentaltier (Horne, 2000). With regard to the second point, relating to writing as parasiticon speech, we should treat the relationship between spoken and writtenlanguage as a hypothesis in order to search for true empirical findings.

Surprisingly, the possibility of circularity is hardly ever discussed inside theparadigm, even if there are strong potentials in such a direction. Goswami andBryant’s (1990) questioning of the phonemic focus is a relativization of thephonemic structure, but this insight is simply added to and rendered harmless inthe paradigm’s phonological theory. Ehri’s (1997) position of written language aslinguistic also strongly challenges the main argument of the current theory.Nevertheless, it is enclosed in the theory as an alternative to Liberman’s view,and no mention is made of its potential for falsification of the mainstream theory.

One may ask how such patience with findings contradicting a theory can bepossible in an empirical science. A plausible answer is that this ‘patience’ is dueto overly vague definitions. Let us take an example from geometry: If we want togeneralize the concept of a circle, we have to set some features of some circlesaside. If we investigate circles of different diameters, we will discover that,independently of the length of the diameter, the circumference will be 3.14 timeslonger. The more general term ‘circle’ does not have a fixed circumference ordiameter, and we therefore disregard these features. This notion of ‘circle’ istherefore more abstract than the notion of ‘a circle whose circumference is 10 cm’.However, if we generalize further from the notion of ‘circle’ to the notion of‘figure’, we obtain a term which is less useful in science. This is because, in thenotion of ‘figure’, we set aside so many features that defining the notion becomesdifficult. As a consequence, the notion of ‘figure’ lacks precision compared with‘circle’, and will also be less fruitful as a scientific notion. Analogously,autonomous linguistics gives priority to an abstract construction which is notdirectly accessible by empirical data, resulting in a situation where the abstractconstruction is confused with the features observed.

If we take a look at the state of the art in dyslexia research, we find thatVellutino et al. (2004) highlight the causal relationship between phonological

awareness and the ability to read and write. Their notion of causality is, however,constrained by what we have claimed to be poor definitions of basic terms. If thenotion of ‘phonology’ lacks precision, then the argument for causality is flawed.

This is to say that, viewed from a different theoretical position, the mainargument used in theories of phonological awareness can be described ascircular. It is therefore, in light of the philosophy of science, high time to shout‘the Emperor has no clothes!’. While such a shout does not amount to aconclusion, it does at least provide a new perspective which may underline thehypothetical character of every paradigm’s phonological theory. Another relatedproblem is the lack of distance to their own theoretical foundation which can beidentified in, for example, Vellutino et al. (2004). The notion of ‘logic’ in thefollowing quotation in fact refers only to the borders of their own theory:‘Etiological theories, which implicate deficits in such abilities as causally relatedto reading difficulties, can be ruled out on logical grounds alone and they havenot fared well in empirical research.’ (Vellutino et al., 2004). And if we go on tocompare the lack of precision of the concepts of ‘phonology’ and ‘linguistic’ withVellutino et al.’s objection to the dual-route model, we see a clear mismatch ofideal and practice: ‘Although such findings are suggestive, the double deficithypothesis can be challenged on theoretical, interpretive, and methodologicalgrounds. As regards its theoretical underpinnings, we suggest that the ‘precisetiming mechanism’ that presumably underlies the formation of orthographiccodes lacks the type of specification that would lend it psychological reality andallow it to be evaluated experimentally as a valid hypothetical construct.’(Vellutino et al., 2004, p. 14).

A spectre is haunting reading research, and especially the field ofdyslexia}the spectre of the phoneme. Interestingly, most theorists reject thisspectral notion in some ways. With reference to experimental research, Vellutinoclaims that the phoneme is not the unit used in speech processing (Vellutino,1979). A.M. Liberman first claimed the phoneme as the unit in speech. Later on,he has repented and admitted that there is no discrete, observable phoneme;instead, it can be found deep in the phonetic module in the shape of phoneticgestures (Liberman, 1997). Goswami and Bryant reject the phoneme as the onlystructure one can be aware of. Nevertheless, the phoneme is still there. And theproblem is that, in the maintenance of the theory of the phoneme, no alternativesto the basic theory have been thoroughly discussed; instead, they aredogmatically rejected (see for example Liberman’s rejection of Tallal andLindblom (Liberman, 1999)). As a consequence, we have a situation where thephoneme is both rejected and accepted, which naturally only enhances theconfusions. The inductive character of theory-building is especially clearly seenin the writings of Alvin and Isabel Liberman, (Liberman, 1997, 1999; Libermanet al., 1989), where dogmatic arguments are deployed against features oftheoretical positions which are not at all compatible with the authors’ ownposition. While this kind of controversy is of course not unusual in science, theproportion of arguments belonging to the dogmatic category is alarming. Thesearguments are clearly not sufficient to prove A.M. Liberman’s claim about therelationship between spoken and written language. In our view, dogmaticpositions should be avoided in order to maintain high standards of empiricalscience. This can be done by studying behaviour in written and spoken language,without a priori assumptions of causal relationships.

THE AUDITIVE-DEFICIT HYPOTHESIS

The position of Tallal and her co-workers is interesting because it partlyrepresents a break with the tradition presented above, and because it places thefocus on perception as a source of deficit. (Even though Tallal’s work waspointing in this new direction already in the 1970s, it is only in recent years thatmore considerable attention has been paid to it.) In this respect, the work of Tallalstands as an alternative to the mainstream phonological focus. Tallal et al. (1996)claim that her earlier work on ‘language-learning-impaired (LLI) children’ hasshown that: ‘[R]ather than deriving from a primarily linguistic or cognitiveimpairment, the phonological and language difficulties of LLI children mayresult from a more basic deficit in processing rapidly changing sensory inputs.’(p. 81). One of the reasons why this research has received a great deal of attentionin recent years is that CD-ROM programs have been developed which are said tobe very effective in treating language difficulties in general and dyslexia inparticular: ‘In the present study we have demonstrated that training childrenwith speech stimuli in which the brief, rapidly changing components have beentemporally prolonged and emphasized, coupled with adaptive training exercisesdesigned to sharpen temporal processing abilities [. . .] results in a dramaticimprovement in receptive speech and language in LLI children.’ (Tallal et al.,1996, p. 83).

In this part of the article, we wish to present and discuss the findings andtheories which have been used to explain the data and the various claims madeby Tallal and her co-workers. This will be done as an alternative to mainstreamthinking, but also to proceed towards conceptual clarity, which we claim cannotbe obtained unless one goes beyond the positions evaluated in the present article.

The notion of ‘dyslexia’ in the auditive-deficit hypothesis.

Miller and Tallal (1996) have not studied dyslexics per se, but rather the broadergroup of what they term ‘language-learning-impaired (LLI) children’. This groupis rather unclearly delimited, as can be seen from this partial definition: ‘We usethe term Specific Language Impairments to refer to a type of communicationdisorder similar to the Diagnostic and Statistical Manual of Mental Disorders}

Fourth Edition (DSM-IV, 1994).’ (note on p. 9). Even though Miller & Tallal (1996,p. 9) remind us that ‘phonology is one of five basic components of oral language(phonology, semantics, morphology, syntax, pragmatics)’, it would seem to be thecase that they consider phonology to be the most important factor of language. Inparticular, it seems as if they are claiming that explaining phonological problemsis much the same thing as explaining most cases of LLI: ‘Taken together, theseexperimental findings led us to hypothesize that the deficits underlying thephonetic reception limitations of a LLI child might arise in early life as aconsequence of abnormal perceptual learning that then contributes to abnormallanguage learning.’ (Merzenich et al., 1996, p. 77). Dyslexics, then, seem to be amajor subgroup of LLI children. The imprecise delimitation of the dyslexia groupmakes it difficult to draw any conclusions regarding samples based on morespecific definitions of ‘dyslexia’ (cf. T�nnessen, 1995, 1997; cf. also Bishop &Snowling, 2004).

SymptomsDuring the last two or three decades, Tallal and her co-workers have claimed thatthe phonological deficit found in dyslexics and dysphasics can be explained as aresult of auditive anomalies (Tallal & Piercy, 1973, 1974; Tallal et al., 1996). Forexample, Tallal (1984) claims that the phonological deficit is a symptom of anunderlying deficit in auditory temporal processing. Tallal, Miller, and Fitch (1993)present three interlocking claims: (1) rapid auditory temporal processing is adecisive factor in speech perception; (2) the specialization of the left hemispherefor speech perception (and phonology) found in the vast majority of right-handed people has to do with this hemisphere’s specialized ability to performrapid auditory temporal processing; and (3) the phonological deficit typicallyfound in, among others, dyslexics is due to a deficit in rapid auditory temporalprocessing.

Moreover, they make the further claim that ‘LLI children commonly cannotidentify fast elements embedded in ongoing speech that have durations in therange of a few tens of milliseconds, a critical time frame over which manyphonetic contrasts are signalled [. . .] For example, LLI children have particulardifficulty in discriminating between many speech syllables, such as [ba] and [da],which are characterized by very rapid frequency changes (formant transitions)that occur during the initial few tens of milliseconds.’ (Tallal et al., 1996, p. 81).These children’s problems in identifying phonetic details and nuances inspeech are said to have many negative consequences: ‘[T]his basic temporalprocessing deficit may disrupt the normal sharpening of neurally representedphonetic prototypes for the native language in LLI children, resulting in acascade of negative effects on subsequent receptive and expressive languagedevelopment}and ultimately resulting in a failure to generate the robustphonetic code that is so essential to learning to read.’ (Tallal et al., 1996, p. 82).

This means, first of all, that we should be finding that dyslexics exhibit areduced ability to understand speech. There are indeed some findings that wouldseem to indicate that this is the case, in particular with regard to the youngestdyslexics. On the other hand, it is often claimed that a characteristic trait ofdyslexics is the discrepancy between their fairly normal ability to understandspeech and their defective understanding of written material (Aaron, 1989;Stanovich & Siegel, 1994; Lyon, 1994).

Another consequence of Tallal’s theory would seem to be that dyslexics shouldexhibit many errors in articulation, and that they should be using words such as‘hat’ and ‘cat’ incorrectly or inconsistently, because they cannot correctly identifyand discriminate between the initial phonemes /h/ and /k/. Now, we do findsymptoms like these in some dyslexics, but rarely to any great degree(Scarborough, 1990).

Tallal et al. (1996) are of the opinion that the ‘temporal-processing deficit’hinders ‘the normal sharpening of neurally represented phonetic prototypes’,thereby leading to a ‘failure to generate the robust phonetic code’. As far as wecan see, this should primarily be causing problems in the young pupil’s grasp ofthe grapheme–phoneme correspondences. If some of the phonemes found in thepupil’s long-term memory are not correct, this will lead to problems in learningto read and write all words containing these phonemes. The child should thenhave equally great problems with these phonemes in frequently occurring wordsand in non-words; what should be decisive is whether or not a word contains the

poorly learned phoneme. This, however, seems to contradict findings whichshow that most dyslexics have most difficulty reading unfamiliar words, pseudo-words and, especially, non-words (cf. Rack, Snowling, & Olson, 1992).

What exactly do Tallal and her co-workers mean when they state that the LLIchild’s phonetic codes are unclear and unstable? Cognitive psychology holds thatthe phonemes are stored as mental representations. It is not immediatelyapparent how a mental representation can be unclear or unstable. Anythingrepresented in long-term memory has to be something definite. (Of course it canalways be wrong, but wrong representations are neither unclear nor unsta-ble}they are merely wrong.)

Does this mean that we are unable to account for ‘unclear’ or ‘unstable’representations within the framework of cognitive psychology? We think not.The best way of doing so, we think, is to be clear about the difference between anability and the exercising of that ability. In other words, we may conceive of therepresentations in long-term memory as being constant, while the performancebased on those representations is variable. The vagaries of the particular situationmay well impact negatively on a person’s performance, but it is difficult to seehow a permanent deficit in the temporal and/or auditive systems couldsometimes have an impact and sometimes not.

When Tallal describes the psychological mechanisms underlying dyslexia, sheoften expresses herself, in our view, unclearly and incompletely. Nonetheless, ourimpression is that her thinking is fairly solidly planted in cognitivism’smentalistic or symbolic paradigm. Her ‘unclear’ or ‘unstable’ mental representa-tions in long-term memory would actually be easier to account for within theconceptual framework of, for example, connectionism. Connectionism is basedon the idea that the activation in a unit and the weights between units arecontinuous rather than binary (0 or l). This allows connectionism to explain alldegrees of unclarity and instability (cf. Bechtel & Abrahamsen, 1991). In our viewcognitive approaches are still predominant in dyslexia research. We considerconnectionism to be a combination of behaviouristic and cognitive approaches(T�nnessen, 1999b). In our view Snowling and Hulme are interesting exceptions(e.g Hulme, Quinlan, Bolt, and Snowling, 1995; Snowling and Hulme, 1999).However, we do not consider that they have used all the potentialities inconnectionism.

This having been said, we can return to the question of why unclear orunstable representations should arise as a consequence of a child not hearingcertain sounds. The sounds that the child does hear, or perceive, are clear andstable; so why should the sounds that he or she does not hear or perceivesuddenly be unclear or unstable?

Regardless of whether we work in a cognitive or a connectionist framework, itis difficult to see how small details and nuances in perception can have such adecisive impact on language learning as Tallal is claiming. Both cognitivism andconnectionism, and other schools as well, stress the role of context in bothperception and learning. It is a generally accepted fact that we usually try to fitinformation together into coherent wholes. Here is an example. Let us say thatsomeone has difficulty in discriminating between the phonemes /h/ and /k/when they appear in isolation. That person would most likely have no troublekeeping things straight when presented with the sentences: ‘Take off your hat’and ‘Let out the cat’. Now, would Tallal claim that our subject in both cases

actually ‘heard’ the word ‘cat’}or the word ‘hat’? Or does our subject hear only/æ t/? When we take into consideration the great amount of redundancy foundin language, it certainly seems quite probable that the /k/–/h/ distinctionproblem would resolve itself over time. After all, a person initially handicappedby not being able to perceive this distinction would be corrected many times inhis or her early years and thus make the necessary adjustments to perception inlight of the context.

Perhaps the most important objections to Tallal’s auditive work have comefrom Isabelle and Alvin Liberman and those working in their tradition.(Shankweiler & Studdert-Kennedy, 1967; Mattingly & Liberman, 1990; Mody,Studdert-Kennedy & Brady, 1997). The question they have raised may beformulated like this: By placing so much emphasis on auditive details and finediscrimination, has Tallal misrepresented the nature of language? Studdert-Kennedy & Mody (1995) comment thus: ‘First, a CV formant transition is not a‘series of rapidly changing acoustic events,’ but an integral spectral sweepreflecting the continuously changing resonances of the vocal tract, as a speakermoves from a point of closure into the following vowel. Second, as manyexperiments have shown (e.g. Mann & Liberman, 1983; Mattingly, Liberman,Syrdal, & Halwes, 1971), a brief formant transition removed from the speechsignal is heard as a rapid, integral glissando, or ‘chirp,’ of which the parts or‘spectral changes’ cannot be perceptually individuated.’ (p. 511).

Mann and Liberman (1983) make this claim: ‘Given transitions that change inrelatively small physical steps, from one appropriate for /d/ to one appropriatefor /g/, the percept changes, not in correspondingly small steps, but suddenly(Mattingly et al., 1971; Studdert-Kennedy & Shankweiler, 1970). This nearlycategorical shift marks a sharp boundary between the phones [d] and [g]; it iscommonly reflected and measured as a relative increase in discriminability of thestimuli at the category boundary.’ (p. 212). In our view, we have enough evidenceto claim that language is categorical in some sense of the word. But we must takecare not to conceive of the boundaries between the phonemes as clear-cut andabsolute. We cannot define phonemes by a set of necessary and sufficientconditions. Instead, the similarities among the sounds categorized as a certainphoneme are rather like the similarities one finds among members of a family.Rosch (1973) measured the reaction time that it took subjects to classify variousinstances under concepts. The finding made was that the degree of typicalitypredicted the speed at which an instance would be classified. Thus a robin, whichrated as a highly typical bird, was quickly classified as a bird, while a goose,which was not thought to be a very typical bird, was classified more slowly.Magnetic attraction is another image used in phonetics to explain how sounds arecategorized into phonemes: the farther away something is from the centre of themagnetic field, the weaker is the influence of the magnet (see e.g. Kuhl, 1991). Fora discussion of the status of the concept of ‘phoneme’, see Uppstad & T�nnessen(2005).

ExplanationsMiller and Tallal (1996) make this claim: ‘[W]e believe that we have identified anunderlying neurological impairment that disrupts temporal processing rates inthe brains of language learning impaired individuals’ (p. 8). Taken together with

the suggestion that the problems dyslexics have can be traced back to anomaliesin the magnocellular system (Livingstone, Rosen, Drislane & Galaburda, 1991;Galaburda, Menard, & Rosen, 1994; Miller, & Tallal, 1996; Fitch, Tallal, Brown,Galaburda, & Rosen, 1994), we discern the following implicit chain of causes inTallal’s and her colleagues’ reasoning: (1) Neurological impairment leading to (2)anomalies in temporal processing leading to (3) auditive anomalies leading to (4)phonological anomalies leading to (5) language and cognitive problems. As wesee their work, only the first link in this chain, the neurological impairment, ispurely biological. All of the subsequent links belong to the domain of psychology.Even though Tallal and her colleagues often give the impression that the causalchain starts off with biological factors, we nonetheless see some statements thatpoint in the opposite direction. For example, Merzenich et al. (1996) write: ‘Thesestudies also strongly indicate to us that there may be no fundamental defect inthe learning machinery in most of these children.’ (p. 80).

If Tallal and her colleagues are claiming that dyslexics’ root problem is alearning problem, then it is difficult to see which type of theory of languagelearning they are basing their assumption on. Theories of language learningwould seem to fall into various types. According to Aslin and Pisoni (1980), thereare four types of theories about the child’s perceptual development: (1) Theuniversal theory claims that children are born with the ability to identify anddiscriminate all sounds in all languages. In the course of the child’s first years oflife, this ‘universal’ ability shrinks, as the phonemes not regularly encountered bythe child are not reinforced. Thus an inborn wealth of nuances and distinctions isgradually reduced. (2) The attunement theory posits a development which goes inthe other direction: here, the child starts out with only a few rough categories,which over time become finer and finer as a result of stimulation. (3) Theperceptual-learning theory assumes that we are not born with any categories at all;everything has to be learned. And finally, (4) the maturational theory says that theability to phonologically identify and discriminate between phonemes unfoldsaccording to a genetically determined time-table, irrespective of early experience.Tallal’s thinking seems to be most in accordance with theory types (2) and (3).

Auditive and temporal factors in dyslexia

Auditive factorsA.M. Liberman and his followers focus on what they see as Tallal’s disregard forthe categorical nature of language. To the extent that one explains phonology onthe basis of non-categorized sound waves, one explains something linguistic onthe basis of something non-linguistic. This may be considered a form ofreductionism (see, e.g. Garfinkel, 1991; Tooley, 1993).

We see a close analogy in the phenomenon of colour. If we try to define colourson the basis of light-wave length, we run into the same problem. While there isindeed a perfect correlation between wavelength and colour, what we have is nota causal relationship. Rather, the relationship between wavelength and perceivedcolour is better described by an image of a mountain top. The mountain has twoor more sides with different appearances. Neither side is the ‘cause’ of theother(s), even though they are of course closely related.

If the theory of Tallal and her co-workers seems reductive, their practice seemsto be the opposite. The training program devised by Tallal and her colleagues

makes one wonder whether they are trying to create something linguistic withthe help of something non-linguistic. Is this possible? Well, first we have toremember that sound waves are a necessary, but not a sufficient, condition forperceiving phonemes. In this sense, we see that auditive factors are relevant tophonology. On purely theoretical grounds, we cannot rule out the possibility thatauditive training can help phonological development. Whether or not this isactually the case is a question that will have to be decided empirically. And it isalso an empirical question to try to determine how detailed the auditiveinformation given to the child should be.

THE NOTION OF ‘PHONOLOGY’ BEYOND COGNITIVISM}THENEED FOR A VULNERABLE THEORY

The problem with A.M. Liberman’s ‘right’ theory is its platonic character, whereevery important phonological feature is beyond falsification. This way ofthinking is very much like Chomsky’s (see Table 1). What is more, awarenessis used for capturing the units of phonology, without any critical discussion ofhow awareness can be defined and operationalized.

In research on dyslexia, we should pick the theory which has the bestdescriptive and explanatory potential, and which is the most vulnerable, i.e.which is easiest to falsify. ‘Vulnerability’ is chosen as a criterion here because itrefers not only to the binary logical property of ‘falsifiability’}i.e. whether ahypothesis can or cannot be falsified. Rather, this notion focuses on the ease withwhich a hypothesis can be falsified, and it is therefore a practical criterion asregards the search for scientific truth. Our notion of ‘vulnerability’ as a practicalcriterion is in line with Karl Popper’s philosophy of science (Popper, 1965;Popper & Eccles, 1977). Indeed, Tallal’s theory is highly vulnerable, and thereforeit can be thoroughly discussed and perhaps falsified. In this respect, Tallal’stheory is more valuable than e.g. Liberman’s position.

In mainstream dyslexia research, phonological representations play a centralrole (Uppstad, 2006a; Uppstad & T�nnessen, 2005). We claim that a strong focuson representations constitutes a major obstacle to a revitalized conception of thenotion of ‘phonology’ beyond phonological awareness. This is because the idea ofrepresentations presupposes introspection, which cannot be falsified. If weconsider the enterprise of science to be about description, explanation andunderstanding, the notion of ‘phonology’ must also be located with reference tothese different purposes. Representations serve the purpose of description andunderstanding. In linguistics, one rarely leaves the domain of description, andthis purpose is also predominant in the linguistic theory that constitutes thefoundation of cognitive sciences. This fact represents a problem when causalhypotheses such as the phonological-deficit hypothesis are presented: it is notpossible to state any falsifiable hypotheses concerning causality. Scientificdisciplines searching for explanations have to be highly sensitive to the qualityof their conceptual apparatus. If we want to explain, we need a conception ofhow descriptions and explanations relate. In our view, the purpose of explanationplaces specific demands on how we elaborate and maintain descriptions. As wesee it, the most fruitful position is to consider descriptions as hypotheses(T�nnessen, 1997). The overall problem with scientific disciplines focusing on

description is that they produce static perspectives. This is the case with bothtraditional linguistics and cognitive psychology. In our view, the connectionistposition seems to represent the best potential for escaping from cognitive scienceand its representational, descriptive focus (cf. Bechtel & Abrahamsen, 1991).Phonology is primarily descriptive, but in reading research it is}without furtherreasoning}widely applied as explanation. Nicolson (2002) points to this problemfor proponents of the phonological-deficit hypothesis. The same problemis associated with the status of the concept of ‘phoneme’ (Uppstad & T�nnessen,2005).

A vulnerable theory also requires precise definitions of core concepts. A moreprecise statement is easier to falsify than a vaguer one. In this sense, the notion of‘phonological awareness’ faces an uncertain future. So far we have focused on thenotion of ‘phonology’, but the other part of the phrase, ‘awareness’, probablyneeds as much rethinking as the first. Important questions in this context are:(1) Are the important aspects of sound seized through awareness? and (2) Isawareness isolated from other skills? In our view, the lack of adequate answers tosuch questions highlights the vague conceptual apparatus found in the theory ofphonological awareness. Concerning the first question, Goswami & Bryant (1990)present interesting positions which relativize the phonemic focus. Still, this is apotential that cannot be fully exploited in the framework of phonologicalawareness and cognitivism. Concerning the second question, T�nnessen (1999a)claims that awareness cannot be separated from automaticity. This positioninvolves a relativization of awareness, in suggesting that awareness andautomaticity can be studied adequately only within the notion of skill.

A direct consequence of the outlined (vulnerable) theory is the adequacy ofstudying linguistic behaviour in written language. By studying reading andwriting behaviour without a priori assumptions about the relationship betweenspoken and written language, we focus on the activity where the problems ofdyslexics become the most visible (see Uppstad, 2006b for a suggestion of howthis can be done). This means approaching dyslexia through a study ofsymptoms e.g. in accordance with psychiatric diagnosing system DSM(T�nnessen, 1997). When searching for the causes of dyslexia in spokenlanguage, we should provide definitions without causal factors which shouldbe treated as hypotheses in daring falsification designs.

CONCLUSIONS

In this article we have focused on the dominant conception of ‘phonology’ incognitive psychology. We have claimed that the theory of phonological awarenesshas a definition of ‘phonology’ which is not suitable to empirical research, owingto its vagueness and empirical status. Further, we have investigated the attemptby Tallal and her co-workers to stretch the notion of ‘phonology’ towardsperceptual and auditory aspects. We have claimed that this approach lacks auseful level of generalization of language structure, and that the criteria ofempirical science have not been met. Definitions of ‘dyslexia’ should be based onsymptoms and should not include causes such as ‘a phonological deficit’(T�nnessen, 1997), because then we delimit the search for causes and risk circularreasoning. E.g. if we include ‘phonological deficit’ in the definition of ‘dyslexia’

and sample dyslexics on the basis of this definition, then it is not surprising thatwe fin a high frequency of phonological deficits among dyslexics. To the extentthat notions such as ‘phoneme’ and ‘phonology’ are included in hypotheses, theirstatus must be clarified and they must be operationalized according to commonlyaccepted standards of empirical science.

In Karl Popper’s concept of ‘falsification’, we find the illustration of theblack swan falsifying the statement ‘all swans are white’. Still, the enterprise offalsification is not as simple as it may seem, because one must verify that theblack swan truly is black and that it truly is a swan. In other words, we need toinvestigate whether deaf people’s reading ability correspond to the exceptionsthat Popper illustrates with the black swans. Faced with this population withinthe literate community, phonology according to any precise, scientific definitioncannot be seen as either necessary or sufficient for the ability to read. Proponentsof the phonological-deficit hypothesis have chosen either to ignore this fact or touse non-empirical theory to explain it away. For the proponents of thephonological-deficit hypothesis, some central challenges must be met: (1) Canthe hypothesis be formulated in such a way that it can be falsified? and 2) Whichempirical findings could falsify the hypothesis?

References

Aaron, P. G. (1989). Dyslexia and hyperlexia. Dordrecht: Kluwer Academic.

Aslin, R. N., & Pisoni, D. B. (1980). Some developmental processes in speech perception.In G. H. Yeni-komshian, J. H. Kavanagh, & C. A. Ferguson (Eds.), Child phonology (Vol. 2:Perception). New York: Academic Press.

Bechtel, W., & Abrahamsen, A. (1991). Connectionism and the mind: An introduction to parallelprocessing in networks. Cambridge, MA: Basil Blackwell.

Bishop, D. V. M., & Snowling, M. (2004). Developmental dyslexia and specific languageimpairment: Same or different? Psychological Bulletin 130(6), 858–886.

Brady, S., Scarborough, H., & Shankweiler, D. (1996). A perspective on two recent researchreports. Perspectives, 22, 5–8.

Chomsky, N. (1957). Syntactic structures. The Hague: Mouton.

Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press.

Chomsky, N. (1975). Reflections on language. New York: Pantheon Books.

Chomsky, N., & Halle, M. (1968). The sound pattern of English. London: Harper and Row.

Churchland, P. S. (1986). Neurophilosophy: Toward a unified science of the mind/brain.Cambridge, MA: MIT Press.

DSM-IV. (1994). Diagnostic and statistical manual of mental disorders (4th ed.).Washington DC: American Psychiatric Association.

Ehri, L. C. (1997). Sight word learning in normal children and dyslexics. In B. B. Blackman(Ed.), Foundations of reading acquisition and dyslexia. Implications for early intervention(pp. 163–190). London: Lawrence Earlbaum Associates.

Fitch, R. H., Tallal, P., Brown, C. P., Galaburda, A. M., & Rosen, G. D. (1994). Inducedmicrogyria and auditory temporal processing in rats: A model for language impairment?Cerebral Cortex, 4(3), 260–270.

Galaburda, A. M., Menard, M. T., & Rosen, G. D. (1994). Evidence for aberrant auditoryanatomy in developmental dyslexia. Proceedings of the National Academy of Sciences, 91,8010–8013.

Gardner, H. (1985). The mind’s new science: A history of the cognitive revolution. New York:Basic Books.

Garfinkel, A. (1991). Reductionism. In R. Boyd, P. Gasper, & J. D. Trout (Eds.),The philosophy of science. Cambridge, MA: MIT Press.

Goswami, U., & Bryant, P. (1990). Phonological skills and learning to read. Hove: Erlbaum.

Gumperz, J. J., & Levinson, S. C. (Eds.). (1996). Rethinking linguistic relativity. Cambridge:Cambridge University Press.

Horne, M. (Ed.). (2000). Prosody: Theory and experiment : studies presented to Gosta Bruce.Dordrecht: Kluwer.

Hulme, C., Quinlan, P., Bolt, G., & Snowling, M. (1995). Building phonological knowledgeinto a connectionist model of the development of word meaning. Language and CognitiveProcesses, 10, 387–391.

Jakobson, R., Taylor, M., & Waugh, L. R. (2002). The sound shape of language (3rd ed.). Berlin:Mouton de Gruyter.

Kager, R. (1999). Optimality theory. Cambridge: Cambridge University Press.

Kuhl, P. K. (1991). Human adults and human infants show a ‘perceptual magnet effect’for the prototypes of speech categories, monkeys do not. Perception and Psychophysics, 50,93–107.

Kuhn, T. S. (1970). The structure of scientific revolutions (2nd enlarged ed.). Chicago:University of Chicago Press.

Liberman, A. M. (1997). How theories of speech affect research in reading and writing.In B. A. Blachman (Ed.), Foundations of reading acquisition and dyslexia: Implications for earlyintervention. Mahwah, NJ: L. Erlbaum Associates.

Liberman, A. M. (1999). The reading researcher and the reading teacher need the righttheory of speech. Scientific studies of reading: The official journal of the Society for the ScientificStudy of Reading, 3(2), 95–113.

Liberman, I. Y., Shankweiler, D., & Liberman, A. M. (1989). The alphabetic principle andlearning to read. In D. Shankweiler, & I. Y. Liberman (Eds.), Phonology and reading disability:Solving the reading puzzle. Ann Arbor: University of Michigan Press.

Linell, P. (1982). The written language bias in linguistics. Linkoping: Universitetet.

Livingstone, M. S., Rosen, G. D., Drislane, F. W., & Galaburda, A. M. (1991). Physiologicaland anatomical evidence for a magnocellular defect in developmental dyslexia. Proceedingsof the National Academy of Sciences, 88, 7943–7947.

Lyon, G. R. (Ed.). (1994). Frames of reference for the assessment of learning disabilities:New views on measurement issues. Baltimore, MD: Paul H. Brookes.

Mann, V. A., & Liberman, A. M. (1983). Some differences between phonetic and auditorymodes of perception. Cognition, 14, 211–235.

Matthews, P. H. (1993). Grammatical theory in the United States from Bloomfield to Chomsky.Cambridge: Cambridge University Press.

Mattingly, I. G., & Liberman, A. M. (1990). Speech and other auditory modules. In G. M.Edelman, W. E. Gall, & W. M. Cowan (Eds.), Signal and sense: Local and global orderin perceptual maps. New York: Wiley.

Mattingly, I. G., Liberman, A. M., Syrdal, A. K., & Halwes, T. (1971). Discrimination inspeech and non-speech modes. Cognitive Psychology, 2, 419–430.

Merzenich, M. M., Jenkins, W. M., Johnston, P., Schreiner, C., Miller, S. L., & Tallal, P. (1996).Temporal processing deficits of language-learning impaired children ameliorated bytraining. Science, 271, 77–81.

Miles, T. R., & Miles, E. (1990). Dyslexia: A hundred years on. Milton Keynes: OpenUniversity Press.

Miller, S., & Tallal, P. (1996). Acoustically modified speech training studies: Implicationsfor dyslexia? Perspectives, 22(3), 8–9.

Mody, M., Studdert-Kennedy, M., & Brady, S. (1997). Speech perception deficits in poorreaders: Auditory processing or phonological coding? Journal of Experimental ChildPsychology, 64, 199–231.

Nicolson, R. I. (2002). The dyslexia ecosystem. Dyslexia: An International Journal of Researchand Practice, 8, 55–66.

Ohman, S. (1979). Introduction: Gunnar Fant, pioneer student of human speech. InS. Ohman, & B. Lindblom (Eds.), Frontiers of speech communication research (pp. xv–xxv).London, New York, San Francisco: Academic Press.

Popper, K. R. (1965). Conjectures and Refutations: The growth of scientific knowledge (2nd ed.).New York: Harper and Row.

Popper, K. R., & Eccles, J. C. (1977). The self and its brain. Berlin: Springer International.

Rack, J. P., Snowling, M. J., & Olson, R. K. (1992). The nonword reading deficit indevelopmental dyslexia: A review. Reading Research Quarterly, 27, 28–53.

Rosch, E. (1973). On the internal structure of perceptual and semantic categories. InT. E. Moore (Ed.), Cognitive development and the acquisition of language. New York: Academic Press.

Scarborough, H. S. (1990). Very early language deficits in dyslexic children. ChildDevelopment, 61, 1728–1743.

Shankweiler, D., & Studdert-Kennedy, M. (1967). Identification of consonants and vowelspresented to left and right ears. Quarterly Journal of Experimental Psychology, 19, 59–63.

Stanovich, K. E., & Siegel, L. S. (1994). Phenotypic performance profile of children withreading disabilities: A regression-based test of the phonological-core variable-differencemodel. Journal of Educational Psychology, 86, 24–53.

Snowling, M., & Hulme, C. (1999). A connectionist perspective on reading and itsdifficulties. In I. Lundberg, F. E. T�nnessen, & I. Austad (Eds.), Dyslexia: Advances in theoryand practice (pp. 45–58). Dordrecht, Boston, London: Kluwer Academic Publishers.

Studdert-Kennedy, M., & Mody, M. (1995). Auditory temporal perception deficits in thereading-impaired: A critical review of the evidence. Psychonomic Bulletin & Review, 2(4),508–514.

Studdert-Kennedy, M., & Shankweiler, D. (1970). Hemispheric specialization for speechperception. Journal of the Acoustical Society of America, 48, 579–594.

Tallal, P. (1984). Temporal or phonetic processing deficit in dyslexia? That is the question.Applied Psycholinguistics, 5, 167–169.

Tallal, P., Miller, S., & Fitch, R. H. (1993). Neurobiological basis of speech: A case forthe preeminence of temporal processing. In P. Tallal, A. M. Galaburda, R. R. Llinas, &C. von Euler (Eds.), Temporal information processing in the nervous system. New York:New York Academy of Sciences.

Tallal, P., Miller, S. L., Bedi, G., Byma, G., Wang, X., Nagarajan, S. S., Schreiner, C.,Jenkins, W. M., & Merzenich, M. M. (1996). Language comprehension in language-learningimpaired children improved with acoustically modified speech. Science, 271, 81–84.

Tallal, P., & Piercy, M. (1973). Developmental aphasia: Impaired rate of nonverbalprocessing as a function of sensory modality. Neuropsychologia, 11, 389–398.

Tallal, P., & Piercy, M. (1974). Developmental aphasia: Rate of auditory processing andselective impairment of consonant perception. Neuropsychologia, 12, 83–93.

Taylor, J. R. (1991). Linguistic categorization: Prototypes in linguistic theory. Oxford: ClarendonPress.

T�nnessen, F. E. (1995). On defining ‘dyslexia’. Scandinavian Journal of Educational Research,39(2), 20.

T�nnessen, F. E. (1997). How can we best define ‘Dyslexia’? Dyslexia, 3, 78–92.

T�nnessen, F. E. (1999a). Awareness and automaticity in reading. In I. Lundberg,F.E. T�nnessen, & I. Austad (Eds.), Dyslexia: Advances in theory and practice (pp. 91–99).Dordrecht, Boston, London: Kluwer Academic Publishers.

T�nnessen, F. E. (1999b). Options and limitations of the cognitive psychological approachto the treatment of dyslexia. Journal of Learning Disabilities, 32(5), 386–393.

Tooley, M. (1993). Causation: Reductionism versus realism. In E. Sosa, & M. Tooley (Eds.),Causation. Oxford: Oxford University Press.

Uppstad, P. H. (2005). Language and literacy: Some fundamental issues in research on readingand writing. Unpublished doctoral thesis, Department of Linguistics and Phonetics, LundUniversity, Lund.

Uppstad, P. H. (2006a). Written language skills and the notion of ‘lexicon’. L1 EducationalStudies in Language and Literature, 6(1). Retrieved December 20, 2005, fromhttp://l1.publication-archive.com/start

Uppstad, P. H. (2006b). The dynamics of written language aquisition. L1 EducationalStudies in Language and Literature, 6(1). Retrieved December 20, 2005, fromhttp://l1.publication-archive.com/start

Uppstad, P. H., & T�nnessen, F. E. (2005). The status of the concept of ‘phoneme’ inpsycholinguistics. In P. H. Uppstad, Language and literacy: Some fundamental issues inresearch on reading and writing. Unpublished doctoral thesis, Department of Linguistics andPhonetics, Lund University, Lund.

Vellutino, F. R. (1979). Dyslexia: theory and research. Cambridge, MA: MIT Press.

Vellutino, F. R., Fletcher, J. M., Snowling, M. J., & Scanlon, D. M. (2004). Specific readingdisability (dyslexia): What have we learned in the past four decades? Journal of ChildPsychology and Psychiatry, 45(1), 2–40.

Wengelin, A. (2002). Text production in adults with reading and writing difficulties. GothenburgMonographs in Linguistics 20. Gothenburg: Departmental of Linguistics, GoteborgUniversity.